# Platform or Solutions - The Central Question ## The company says that it has a platform, but the deployments show solutions The company tells this story. The company builds a shared platform layer one time. The platform layer has an agent builder, a memory engine, enterprise connectors and governance. The company adds a thin domain layer for each vertical. Thus each new customer is mostly configuration and starts operation quickly. The data show something different. The company has deployments and the customers pay. But the company builds each deployment for one customer. Frequently the deployment is on the premises of the customer. Each deployment has a large quantity of engineering work. The deployments are in many sub-verticals. > [!important] The question for all other points > **Do the deployments share the same platform layer?** If yes, margins and speed increase with each customer. If no, the company is a good vertical-AI services business. An investor must value it as a services business. ## What a technical analysis found - The company is **not a thin wrapper**. It has a coherent layered architecture. It has a classical machine-learning function. It has deployments that give revenue. - The company assembled most of the platform layer from **mainstream components and open-source components**, see [[The Four Layer Stack and What Is Commodity]]. - The figures of the company **show that the claim "deploys in days" is incorrect**, see [[Claims Versus Evidence]]. - A large part of the revenue comes from **services**, see [[Quality of Revenue]]. > [!tip] The best metric for the diagnosis > The metric is the work for the deployment of each customer. The company measures it at different times. The metric has two parts. The first part is the number of engineer-days to the start of operation. The second part is the share of commits in shared code compared with the share in customer-specific code. If the curve goes down, the platform claim is correct. The vault uses the same test in other notes, see [[Deployment Velocity]] and [[Consultancy-to-Platform Transition]]. ## Why it matters The answer decides the valuation. If an investor uses a platform multiple for revenue that comes mostly from services, the calculated result is more than the correct result. The expected-return model for a deal like this needs two estimates. The first estimate uses software multiples, if the deployments use the same platform layer again. The second estimate uses services multiples, if the deployments do not use it again. ## Related - [[Foreman MOC]] - [[Delivery Economics and the Scaling Ceiling]] - [[Domain Packs and the Vertical Playbook]] - [[Industrial AI Unit Economics]] - [[AI Eats Services Not Software]]